A VP I worked with measured his AI rollout by training completion rates. Eighty-three percent of his team had finished the module. He called that a win, and in his next board update, he described adoption as "ahead of schedule."
When I asked what his team actually thought about using the tools in their daily work, he said the feedback had been "mostly positive." I asked if he had heard any real concerns. He paused. "They know this is the direction we're headed," he said.
That phrase is one of the most reliable early-warning signs in organizational change. It means the leader has confused announcement with acceptance. They have interpreted the absence of visible resistance as readiness. They are watching a deployment metric while something far more consequential plays out underneath it.
What the Data Is Actually Showing
HBR published research this spring showing that executives overwhelmingly rate their AI adoption as successful, while most of their teams describe feeling anxious and underprepared. Forbes and Deloitte have reported similar findings across industries. Organizations are spending significantly on training programs, certification pathways, and structured onboarding resources. Completion rates look excellent. Behavioral change does not follow.
This pattern has a name in organizational research: the adoption-performance gap. Tools get installed. Dashboards show usage. The actual work processes barely shift. The organization has technically adopted the technology while functionally resisting it. And the executive sponsors the next round of investment based on metrics that measure presence, not change.
What Leaders Assume the Problem Is
The default executive interpretation of a low-adoption scenario is a capability problem. If people are not using the tools effectively, they must not understand them well enough. The fix, in this framing, is always more training. Better onboarding. A champion network. A lunch-and-learn series. These interventions are not wrong, exactly. They address the wrong bottleneck.
The assumption underneath this framing is that workers are essentially ready to change and just need the technical knowledge to do so. That assumption is rarely examined and is almost always incorrect in the first twelve months of a major AI initiative.
What Is Actually Happening in the Room
When a worker encounters AI tools that touch the core of their function, they are not just learning software. They are running a private, unspoken calculation about their own professional value. The skills they have spent years developing — their institutional knowledge, their judgment, their hard-won expertise in how this specific organization actually works — are now in comparison with something that can approximate many of those things at scale and speed. That comparison does not resolve because someone sends them a training link.
It resolves when they find an answer to a question they are rarely asked out loud: where do I fit in what is coming? Some people find that answer quickly, usually because they have already absorbed enough context to believe their value survives the transition. Others need to see it play out in practice before they can believe it. And some never fully resolve it — which is why they appear compliant in training while functionally doing the minimum in their actual workflow.
The workers slowest to adopt are not the ones who do not understand the technology. They are the ones who understand it too well — who can see exactly where it is heading and have not yet found a way to believe their own value survives the trip.
Executives who are genuinely enthusiastic about AI have already done their own version of this reckoning. They see the technology as leverage. They have found their position inside the new reality. What they often forget is that this process happened for them too — slowly, privately, over months, usually without any formal acknowledgment that it was a process at all. They arrived at enthusiasm. They do not always remember the route.
What Actually Closes the Gap
Leaders who have navigated large-scale organizational change before — military transitions, post-merger integrations, operational turnarounds — tend to recognize surface compliance immediately for what it is. Low usage is not insubordination. Quiet participation in training is not readiness. These are external signals of an internal process that has not finished. Trying to accelerate that process with more training is the organizational equivalent of trying to accelerate grief with a PowerPoint.
What works is creating conditions for meaning-making, not just tool adoption. That means visible conversations about what work will look like in practice — not just what the tools can do in a demo. It means leaders at every level demonstrating their own process of figuring it out, including the parts that are not resolved yet. And it means building the organizational capacity to distinguish between someone still in process and someone who has quietly decided against it. Most organizations cannot make that distinction. Most executives are not asking the question that would help them see it.
The board update will show adoption ahead of schedule. The training completion rates will look good. The implementation partner will call it a success. Six months later, someone will commission a study to figure out why performance has not improved.
If you are running an AI initiative or sitting close to one, the most useful question you can ask right now is not "how do we improve adoption?" It is: what are your people actually working through? I am happy to talk through what you are seeing.
